BUG: fix validate(sample=x) for pl.DataFrame

[m-richards]

Fixes #1912.

I had to feed through some additional information which meant passing through an additional argument to validate or subsample, which differs from the base class implementations. It seemed better to do that in subsample which is less user facing.

Not sure if I found the best place to put tests.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 93.53%. Comparing base (812b2a8) to head (ba4896e).
Report is 199 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1923      +/-   ##
==========================================
- Coverage   94.28%   93.53%   -0.76%     
==========================================
  Files          91      121      +30     
  Lines        7013     9382    +2369     
==========================================
+ Hits         6612     8775    +2163     
- Misses        401      607     +206     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[cosmicBboy]

The backend only accepting LazyFrame was an intentional choice here, to force the backend only use polar's lazy api when implementing the actual validation logic.

Thinking about this a little bit, instead of introducing an additional argument to the subsample api, what if we did the following:

• Preserve the type of polars.DataFrame and pass it into the backend validate method
• Do the subsampling on polars.DataFrame if that's what the user passed in
• Convert it to a lazyframe.
• At the end of the backend validate method, we may also want to convert that back into a polars.DataFrame if that's what the user passed in.

[m-richards]

I'll have a go and see how this works, I was aware that LazyFrame input was deliberate and wasn't sure if it made sense to change that.